In this article I construe Russell's definite description notation as a fragment of an "ideal language"– a language in which, as Russell puts it in the "Logical Atomism" lectures, "the words in a proposition correspond one by one with the components of the corresponding fact." Russell's notation – containing as it does variables, quantifiers and the identity sign – commits him to an ontology that is lavish indeed. It thus conflicts with the spirit of the theory of descriptions, which is (...) developed in the service of ontological frugality. I make use of arguments derived from the Tractatus to show that an ideal language need not contain logical signs. I thus defend the spirit of the theory of descriptions while departing from its letter. (shrink)